Tethered unmanned aerial vehicle air-ground integrated intelligent operation system and method

Abstract

This invention discloses an integrated air-ground intelligent operation system and method for tethered unmanned aerial vehicles (UAVs). The system includes a UAV terminal, a ground station, and a tethered cable connecting the two. The UAV terminal is equipped with an onboard controller, and the ground station is equipped with a ground controller, forming a distributed model prediction control unit. The UAV terminal also has a power state observer, used to calculate a power estimate based on collected voltage and current signals combined with cable resistance parameters. This power estimate is sent to the ground station via a timestamped communication link for the ground controller to perform global optimization calculations. The system also includes a multi-sensor fusion positioning module, a dynamic resource allocator, a primary / backup dual-link communication architecture, and a Lyapunov optimizer. This invention achieves precise control, maximized energy efficiency, and high-reliability operation of tethered UAVs in dynamic environments through distributed control, precise energy sensing, intelligent state estimation, and cross-layer collaborative optimization.

Description

Technical Field

[0001] This invention belongs to the field of tethered unmanned aerial vehicle (UAV) persistent operation and dynamic environment adaptive control technology, and relates to a tethered UAV air-ground integrated intelligent operation system and method. Background Technology

[0002] Tethered unmanned aerial vehicle (UAV) systems provide continuous power and data transmission via cables, offering unique advantages in prolonged operations. However, existing systems still face numerous technical limitations in achieving intelligent and highly reliable operation. Traditional control systems often employ a centralized architecture, making it difficult to simultaneously meet the dual demands of rapid dynamic response from UAVs and global optimization for complex tasks. A single control cycle setting cannot reconcile the conflict between limited onboard computing resources and complex ground optimization, resulting in system response delays or insufficient control precision.

[0003] In terms of energy management, existing systems lack real-time sensing capabilities for tethered cable transmission losses and cannot adaptively adjust based on actual power supply conditions. Power instability is prone to occur when cable impedance changes or power supply fluctuates. The state estimation stage typically employs conventional filtering algorithms, which have limited effectiveness in suppressing accumulated sensor errors and fail to fully utilize environmental disturbance information. Positioning accuracy significantly decreases when GPS signals are limited or wind disturbances are severe, affecting flight stability. Communication systems often rely on a single wireless link, resulting in insufficient reliability in complex electromagnetic environments. Although some systems employ redundant designs, they lack a rapid switching mechanism based on transmission delay monitoring, making it difficult to meet high real-time control requirements.

[0004] Furthermore, the existing system lacks an effective cross-layer coordination mechanism, and the airborne and ground controllers have failed to achieve dynamic coordination based on global state, which restricts the overall performance of the system in complex environments.

[0005] Therefore, there is a need for an integrated air-to-ground solution for tethered drones that can achieve precise control, intelligent perception, and reliable collaboration in order to improve the system's operational efficiency in dynamic environments. Summary of the Invention

[0006] To address the problems existing in the background technology, this invention proposes an integrated air-ground intelligent operation system and method for tethered unmanned aerial vehicles (UAVs).

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a tethered unmanned aerial vehicle (UAV) integrated air-ground intelligent operation system, comprising a UAV terminal, a ground station, and a tethered cable connecting the UAV terminal and the ground station; The UAV terminal is equipped with an onboard controller, and the ground station is equipped with a ground controller; The multi-sensor fusion positioning module, installed on the drone, is used to provide the status information required for the drone's flight. Based on the fact that the control cycle of the airborne controller is shorter than that of the ground controller, a distributed model prediction control unit is constructed. The UAV terminal is also equipped with a power status observer, which is used to calculate the power estimate based on the collected voltage and current signals of the mooring cable and the resistance parameters of the mooring cable. The airborne controller performs rapid response calculations based on the status information to generate a first control command, and the ground controller performs global optimization calculations based on the received power estimate and status information to generate a second control command. The first and second control commands are fused to generate and issue a fused control command to drive the UAV to fly.

[0008] Specifically, the model prediction control unit is configured as follows: Compressed prediction and control time domains are employed; A differentiated weight matrix is ​​used to perform weighted optimization of the system state vector error and control quantity; The weight assigned to position state error is greater than the weight assigned to attitude state error.

[0009] Specifically, a tethered unmanned aerial vehicle (UAV) integrated air-ground intelligent operation system also includes a multi-sensor fusion positioning module, which uses a filtering estimation algorithm to fuse inertial measurement data, satellite positioning data, and environmental wind speed data to obtain the state information containing pose information and environmental disturbance data. The filtering estimation algorithm is configured as follows: The inertial measurement data is used as the basis for system state prediction; The position and altitude of the UAV are observed and updated using the satellite positioning data. The environmental wind speed data is used as an external disturbance input to compensate for the system state prediction.

[0010] Specifically, the filtering estimation algorithm uses a pre-calibrated process noise covariance matrix and an observation noise covariance matrix, both of which are diagonal matrices determined based on sensor performance.

[0011] Specifically, a tethered unmanned aerial vehicle (UAV) integrated air-ground intelligent operation system also includes a dynamic resource allocator, configured as follows: Real-time calculation of the instantaneous fluctuation range of power estimates; When the fluctuation amplitude exceeds the preset safety threshold, the upper limit of the motor speed command in the fusion control command is adaptively lowered.

[0012] Specifically, the communication link adopts a primary and backup dual-link architecture: The main link is a wireless communication link; The backup link is a fiber optic communication link configured using a tethered cable. Seamless switching between primary and backup links is achieved through an automatic switching module.

[0013] Specifically, a tethered unmanned aerial vehicle (UAV) integrated air-ground intelligent operation system also includes a Lyapunov optimizer, configured as follows: Construct a joint optimization algorithm for system stability constraints and system performance objectives, wherein the stability constraints quantify the cumulative deviation of the system through one or more system state queues; The system performance targets are used to characterize the core indicators that the system needs to optimize, including trajectory tracking accuracy, task execution efficiency, and energy consumption economy. Based on the power estimate obtained from the power state observer and the real-time length of the system state queue, the weight parameters in the optimization objective function of the model prediction control unit of the airborne controller and the ground controller are dynamically adjusted. By solving an optimization problem in each control cycle that aims to minimize the weighted sum of the Lyapunov drift term and the system performance penalty term, the long-term average performance and instantaneous stability of the system are balanced in a dynamic environment.

[0014] Based on the above-mentioned tethered UAV integrated air-to-ground intelligent operation system, this technical solution also provides a tethered UAV integrated air-to-ground intelligent operation method, including the following steps: The power state observer collects the voltage and current signals of the moored cable in real time and calculates the power estimate. The multi-sensor fusion positioning module acquires the drone's pose information and environmental disturbance data in real time. Based on the power estimate, pose information and environmental disturbance data, the airborne controller generates a first control command in a first control cycle. Based on the power estimate, pose information and environmental disturbance data, the ground controller generates a second control command in a second control cycle, wherein the first control cycle is shorter than the second control cycle. The first control command and the second control command are fused together to generate a fused control command to drive the UAV to fly.

[0015] Specifically, a method for integrated air-to-ground intelligent operation of tethered unmanned aerial vehicles (UAVs) also includes fault-tolerant control steps: Real-time monitoring of the transmission delay of the communication link used for timestamp synchronization; When the transmission delay exceeds the preset delay threshold, the drone terminal switches to local capacity control mode and triggers an alarm signal.

[0016] Specifically, the step of fusing the first control command and the second control command includes: The joint optimization algorithm constructed using the Lyapunov optimizer dynamically adjusts the fusion weights of the first and second control commands based on the real-time fluctuations of the power estimate. When the fluctuation range of the power estimate is within the first preset range, the weight ratio of the second control command is increased; When the fluctuation range of the power estimate is within the second preset range, the weight ratio of the first control command is increased.

[0017] Compared with the prior art, the present invention has the following beneficial effects: by constructing a distributed model prediction control unit, the system simultaneously possesses the millisecond-level fast response capability of the airborne controller and the global optimization capability of the ground controller, effectively overcoming the contradiction between real-time performance and global capability in the traditional single control unit.

[0018] Based on the precise energy sensing of the power state observer, the adaptive adjustment of power fluctuations by the dynamic resource allocator, and the seamless communication switching between the primary and backup dual links, a comprehensive guarantee mechanism covering multiple dimensions such as energy, execution, and communication is jointly constructed.

[0019] Through the global coordination of the Lyapunov optimizer and the dynamic adjustment of the intelligent instruction fusion strategy, the system achieves an optimal balance between stability and performance, demonstrating excellent environmental adaptability and task execution capabilities.

[0020] These three breakthroughs have enhanced the complexity, reliability, and intelligence of tethered unmanned aerial vehicle (UAV) systems, providing a complete technical solution for persistent intelligent operations. Attached Figure Description

[0021] Figure 1 This is a flowchart of an intelligent air-ground integrated operation method for tethered unmanned aerial vehicles (UAVs) according to the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] like Figure 1 As shown, the technical solution adopted by the present invention is as follows: a tethered unmanned aerial vehicle (UAV) integrated air-ground intelligent operation system, including a UAV terminal, a ground station, and a tethered cable connecting the UAV terminal and the ground station.

[0024] The UAV terminal is equipped with an onboard controller, and the ground station is equipped with a ground controller; The multi-sensor fusion positioning module, installed on the drone, is used to provide the status information required for drone flight.

[0025] A distributed model prediction control unit is constructed based on the fact that the control cycle of the airborne controller is shorter than that of the ground controller.

[0026] The UAV terminal is also equipped with a power status observer, which is used to calculate the power estimate based on the collected voltage and current signals of the mooring cable and the resistance parameters of the mooring cable.

[0027] The airborne controller performs rapid response calculations based on the status information to generate a first control command, and the ground controller performs global optimization calculations based on the received power estimate and status information to generate a second control command. The first and second control commands are fused to generate and issue a fused control command to drive the UAV to fly.

[0028] Furthermore, the drone serves as the operational carrier, carrying the components required for flight control and data acquisition, and is the receiver and execution end for real-time fusion control commands.

[0029] As the overall optimization and decision-making center, the ground station is responsible for coordinating the overall operation status of the system and implementing control planning, but does not directly participate in real-time flight attitude adjustment.

[0030] The tethered cable is the physical connection between the UAV and the ground station. It provides the energy transmission channel required for the continuous operation of the UAV and also serves as a backup carrier for the communication link, ensuring the stability of the connection between the UAV and the ground station.

[0031] The key to the distributed model predictive control unit lies in the design of the control cycle difference between the airborne controller and the ground controller, so as to achieve coordinated control with real-time response and global optimization.

[0032] The airborne controller has a shorter control cycle, enabling it to quickly capture the real-time flight status of the UAV (such as attitude shift and position deviation) and generate targeted first control commands within milliseconds. This meets the real-time control requirements in dynamic environments and avoids flight instability caused by control delays.

[0033] The ground controller has a relatively long control cycle and does not need to focus on instantaneous state adjustments. Instead, it performs global optimization calculations based on more comprehensive system operation data to generate a second control command, ensuring that the control strategy meets the overall operational objectives, such as trajectory planning and energy conservation.

[0034] By employing a distributed division of labor between airborne and ground controllers—namely, short-cycle real-time control and long-cycle global optimization—the problem of a single controller being unable to handle everything is solved. Furthermore, the algorithmic characteristics of model predictive control enable a balance between control accuracy and global efficiency.

[0035] The power state observer is the core component of the system's energy state perception, and its operating logic directly contributes to the accuracy of control optimization. The power state observer is installed at the UAV end to ensure close-range acquisition of real-time electrical signals from the tethered cable, avoiding signal attenuation or interference during transmission. The power state observer specifically acquires the voltage and current signals of the tethered cable, as these two types of signals are core parameters reflecting the energy transmission status. Simultaneously, combined with the inherent resistance parameters of the tethered cable (preset to a known fixed value for the system), a power estimate is calculated.

[0036] The parameter directly acquired is the voltage signal of the mooring cable. With current signal These are the basic electrical parameters for calculating power. The preset fixed parameters are the resistance parameters of the mooring cable. Because cables have resistive losses (especially during long-distance transmission), calculating the apparent power solely by multiplying voltage and current only yields a value. To obtain an accurate estimate of the actual power (i.e., the power estimate), resistance parameters must be considered to correct for losses. The calculation logic can be expressed by the formula: ;in This formula accurately eliminates the effects of cable resistance loss, ensuring that the power estimate reflects the true energy transfer status of the moored system. This formula is applicable to DC power supply systems; for AC power supply systems, power factor compensation is required.

[0037] The power estimate quantifies the energy transmission efficiency and fluctuation status of the moored cable, providing data support for the subsequent global optimization of the ground controller and avoiding mismatch between control commands and actual energy supply due to unknown energy status.

[0038] A key design feature of the communication link is timestamp synchronization, which ensures the accuracy and timeliness of data transmission and command interaction. The timestamp synchronization function precisely timestamps each set of transmitted power estimates, ensuring that the data received by the ground controller strictly corresponds to the real-time status of the UAV, avoiding data lag caused by transmission delays, and ensuring that global optimization calculations are based on the true current state.

[0039] First, the power estimate from the UAV is uploaded to the ground station via this synchronous communication link. The ground controller uses this data, combined with mission requirements, to perform global optimization calculations and generate a corresponding second control command. This second control command is then fused with the first control command generated by the airborne controller to generate a fused control command. Finally, the fused control command is transmitted back to the UAV via the same communication link, driving the UAV to adjust its flight status.

[0040] Specifically, the model prediction control unit of the airborne controller and the ground controller is configured as follows: A compressed prediction and control time domain is used.

[0041] A differentiated weight matrix is ​​used to perform weighted optimization of the system state vector error and control quantity.

[0042] The weight assigned to position state error is greater than the weight assigned to attitude state error.

[0043] Furthermore, in Model Predictive Control (MPC) algorithms, the prediction time domain refers to the number of time steps the algorithm takes to predict the future system state, while the control time domain refers to the number of time steps the algorithm needs to optimize and generate control commands. Compression means shortening the length of these two time domains, rather than using full-time-domain prediction and optimization.

[0044] Airborne controllers are embedded platforms (such as ARM-based SoCs), with limited computing power and the need to meet the high real-time requirements of short control cycles (such as 10ms). If the prediction or control time domain is too long, the number of linear or nonlinear equations that need to be solved in each cycle will increase significantly, causing the computation time to exceed the control cycle and resulting in control lag.

[0045] While ground controllers have greater computing power (such as industrial PCs), their control cycles are longer (such as 100ms). Their goal is global optimization (such as path planning and energy scheduling), rather than instantaneous state tracking. Excessive time domain introduces more future uncertainties (such as wind speed fluctuations and power supply changes), leading to a greater deviation between the optimization results and actual operating conditions. Furthermore, redundant calculations waste computing resources.

[0046] By focusing on recent key states and control variables, time-domain compression reduces the amount of computation per cycle while ensuring that the prediction accuracy meets the system requirements, making the algorithm adaptable to the hardware computing power and cycle requirements of airborne and ground controllers.

[0047] The airborne controller is required to ensure rapid calculation within a short control cycle and to promptly output the first control command to respond to instantaneous changes in the UAV's attitude and position, thereby ensuring flight stability.

[0048] For the ground controller, ensure a balance between global optimization accuracy and computational efficiency, avoid optimization lag or deviation caused by excessive time domain, and ensure that the second control command can fit the actual working conditions.

[0049] A differentiated weighting matrix is ​​used to optimize the system state vector error and control input. This configuration defines two differentiated weighting matrices: a state error weighting matrix and a control input weighting matrix. Control weight matrix The optimization objective of the Model Predictive Control (MPC) algorithm is transformed into minimizing the sum of the weighted state error and the control cost, thereby enabling priority differentiation of different control objectives.

[0050] The objective function of the Model Predictive Control (MPC) algorithm is a quadratic function, in the following form: The optimization objective function focuses on minimizing state error and rationalizing control output, and its mathematical form is: ; in, To optimize the objective function value, a comprehensive index needs to be minimized; for The system state vector at time t; for The system reference state vector at time t; for The transpose of the time-state error vector; for Transpose of the control vector at any moment; for Control vector at any given time; , The time-domain parameters are used to define the prediction and control range.

[0051] Let be the state error weight matrix, a positive semi-definite diagonal matrix (only diagonal elements are non-zero, off-diagonal elements are 0), with diagonal elements... Corresponding to the The weights of each state error.

[0052] The control weight matrix is ​​a positive definite diagonal matrix (all diagonal elements are positive, and off-diagonal elements are 0). The diagonal elements... Corresponding to the The weight of each control quantity.

[0053] Furthermore, The step is the prediction time domain, which refers to predicting the system state for the next 5 sampling periods based on the current state. For example, predicting the position and attitude changes for the next 50 ms under the 10 ms sampling period of an airborne controller is the key to overcoming the computing power limitations of embedded platforms and shortening the single solution time.

[0054] The first step is the control time domain, which means that the execution command is generated only for the first two sampling periods of the prediction. For example, the airborne controller generates a motor speed command of 20ms. The remaining three steps are used as a safety boundary to avoid command failure caused by future state uncertainties (such as sudden gusts of wind), thus balancing foresight and reliability.

[0055] A differentiated weighting matrix is ​​used to optimize the system state vector error and control input using weighted averages. The differentiation is reflected in... Matrix and The diagonal elements of the matrix have different values.

[0056] For states that require priority control (such as location in a work scenario), increase their corresponding... diagonal elements of a matrix . The larger the value, the higher the proportion of the state error in the objective function value, and the more preferentially the algorithm will reduce the state error during optimization.

[0057] For control variables that need to suppress excessive output (such as avoiding sudden increases or decreases in motor speed), increase their corresponding values. diagonal elements of a matrix . The larger the value, the higher the proportion of the control quantity in the objective function value. When optimizing the algorithm, it will tend to output a smoother control quantity with a smaller amplitude to avoid wear and tear on the actuator (motor) or wasted power.

[0058] Matrix diagonalization is designed to simplify computational complexity, avoid cross-coupling effects between different state errors and different control variables, and ensure the targeted nature of weight adjustments.

[0059] This differentiation allows for prioritization of multiple control objectives. For example, in energy-scarce scenarios, it can increase... The weights of motor power-related control variables in the matrix are used to suppress excessive power consumption. Matrix weight constraints prevent drastic fluctuations in control quantities, reduce motor start-stop frequency and overload risk, extend equipment life, and lower energy consumption. In precision operation scenarios, this can increase... The weights of positional errors in the matrix are prioritized to ensure operational accuracy.

[0060] The configuration that assigns a greater weight to position state errors than to attitude state errors is a concrete implementation of the "differentiated weight matrix." It adjusts the weights of the position and attitude states in the system state vector within the state error weight matrix. The corresponding weights in the formula ensure that position control has a higher priority than attitude control.

[0061] System state vector composition: ,in For drones , , The positional state of the direction. Roll angle, Pitch angle, The yaw angle represents the attitude state of the UAV. The system state vector is a set of core parameters describing the current operating state of the UAV. As a state error weight matrix, the number and order of its diagonal elements are perfectly aligned with the components of the system state vector, forming a one-to-one structural mapping relationship.

[0062] The specific settings of the matrix are as follows ,in These are the weighting coefficients for the position state error. The weighting coefficients for attitude state errors are significantly higher than those for position state errors. This is because when there is a conflict between position and attitude state control, such as when precise position path tracking requires temporarily increasing the attitude angle deviation, the objective function is optimized accordingly, given the much larger weighting coefficients for position and attitude errors. The contribution of position error is higher in the middle, so the algorithm will prioritize reducing position error and allow for short-term, small deviations in attitude state.

[0063] In core applications of tethered drones, such as power line inspection, terrain mapping, and targeted material delivery, positional status directly determines the quality of operations. For example, during inspections, a fixed distance from the power line must be maintained, and during delivery, precise placement is crucial. Short-term deviations in attitude, such as slight tilting, do not significantly affect operational results. This configuration ensures the achievement of core operational objectives. It also avoids instability caused by excessive attitude adjustments. If attitude weight is too high, the controller may sacrifice positional accuracy to maintain attitude stability, causing the drone to deviate from the operational path. This configuration clearly prioritizes and balances positional accuracy and attitude stability, preventing control malfunctions caused by prioritization confusion.

[0064] Specifically, the multi-sensor fusion positioning module uses a filtering estimation algorithm to fuse inertial measurement data, satellite positioning data, and environmental wind speed data to obtain the state information that includes pose information and environmental disturbance data.

[0065] The filtering estimation algorithm is configured as follows: The inertial measurement data is used as the basis for system state prediction.

[0066] The position and altitude of the UAV are observed and updated using the satellite positioning data.

[0067] The environmental wind speed data is used as an external disturbance input to compensate for the system state prediction.

[0068] Furthermore, the multi-sensor fusion positioning module integrates inertial measurement data, satellite positioning data, and environmental wind speed data through a filtering estimation algorithm to compensate for the performance limitations of a single sensor. Ultimately, it achieves high-precision, high-real-time, and high-anti-interference estimation of the UAV's pose and motion state, providing reliable state feedback for generating the first and second control commands.

[0069] Filtering estimation algorithm configuration one: using inertial measurement data as the basis for system state prediction.

[0070] This configuration leverages the high-frequency continuity of inertial measurement data to provide real-time, continuous initial predictions of the system state, avoiding state update interruptions caused by sampling gaps with other sensors.

[0071] Inertial measurement data comes from an inertial measurement unit (IMU) and specifically includes two core types of data: Angular velocity measurement: UAV roll angle Pitch angle Yaw angle The real-time rate of change. The linear acceleration measurement value is the UAV in the geodetic coordinate system. , , Real-time acceleration in three directions.

[0072] The sampling frequency is extremely high (usually 100-1000Hz), which can capture the instantaneous motion changes of the drone, but there is a cumulative drift error. For example, the acceleration measurement is affected by the bias, and long-term integration will cause the position or velocity deviation to gradually increase.

[0073] To comprehensively describe the motion state of the UAV and suppress IMU drift, the system state vector of the filtering algorithm... Defined as: ; in: This refers to the position and status of the UAV in the geodetic coordinate system. The velocity state of the UAV in the geodetic coordinate system; The attitude angles of the UAV are roll, pitch, and yaw; The bias error of the inertial measurement unit's accelerometer sensor changes slowly over time and needs to be estimated in real time to compensate for the drift. Similarly, the bias error of the angular velocity sensor in the inertial measurement unit needs to be estimated in real time.

[0074] It should be added that, among them and Both are system state vectors. It is a multi-dimensional, principle-level general vector without explicit time constraints, containing position, velocity, attitude, and acceleration, used to fully describe the core state and dynamic model of the algorithm; while It is a binding discrete time step The 6-dimensional engineering calculation vectors that retain only position and attitude are the core subset of the former and have specific values. This is to adapt to the iterative calculation requirements of the model prediction control unit. By eliminating non-core velocity components, the computing power burden of the embedded platform is reduced and the real-time control is guaranteed. At the same time, the core operational requirements of position accuracy and attitude stability are focused on. It is also compatible with the division of labor between the airborne controller and the ground controller for global coordination and local response, so as to achieve a balance between the integrity of the principle and the practicality of engineering.

[0075] A nonlinear state transition function is constructed based on the Newton-Euler dynamics equations. Using high-frequency measurements from an inertial measurement unit (IMU), The optimal state estimate at time t is derived as follows: The core formula for predicting the state at time t is: ; for The predicted state vector at time step; : The optimal state estimate updated after each observation, representing the most accurate historical state; : The measurements from the inertial measurement unit at any given time (angular velocity measurements, sensor characteristics) serve as the input driving the state transition; function The physical meaning is to transform the changes in motion measured by the inertial measurement unit into absolute quantities of state through the laws of dynamics, for example: velocity update: , This represents the integral after deducting the acceleration bias. This is the sampling interval for the inertial measurement unit. Position update: Position is obtained by integrating velocity. Attitude is updated using angular velocity. After subtracting the bias, the integral yields the attitude angle. The predicted value.

[0076] Nonlinear state transition function It provides continuous high-frequency state prediction. This fills the sampling gaps of low-frequency sensors such as satellite positioning, ensuring that the filtering estimation algorithm outputs a state every millisecond, meeting the real-time requirements of short-cycle control by the airborne controller. It also provides an initial benchmark for subsequent observation updates. Without continuous prediction from the inertial measurement unit, there would be no accurate historical state to correct when satellite positioning data arrives, leading to a break in state estimation.

[0077] Filtering estimation algorithm configuration two: Use satellite positioning data to update position and altitude observations. This configuration is the update step of the filtering algorithm. It uses the absolute accuracy of satellite positioning data to correct the cumulative drift error predicted by the inertial measurement unit, ensuring the long-term positioning accuracy of the UAV's position and altitude.

[0078] Satellite positioning data comes from satellite positioning modules (such as GPS and BeiDou), and specifically includes: Horizontal position data: latitude and longitude coordinates converted to geodetic coordinate system (Eastward) (Northward). Height data: in geodetic coordinate system. (Vertical height);

[0079] The sampling frequency is low (1-10Hz), but there is no cumulative error. It provides absolute positioning accuracy and is used as a reference anchor point to correct drift of the inertial measurement unit.

[0080] The update step of the filtering estimation algorithm corrects the uncertainty of the predicted state by addressing the deviation between the observed and predicted values, and uses the absolute accuracy of satellite positioning to bring back the predicted state that has drifted from the inertial measurement unit. Specific formulas and parameters include: observation model, Kalman gain, state and covariance updates.

[0081] Observation model: Establishes a mapping between states and observations. Satellite positioning can only measure position, therefore the observation model function... Extracting only the position component from the predicted state, the formula is:

[0082] ; for The expected observation at time (dimension 3×1), i.e. ; The observation function is essentially a selection matrix, retaining only the predicted state vector. middle For the corresponding row, set the other dimensions to 0 to ensure that the predicted state matches the observation dimensions of satellite positioning.

[0083] Kalman gain: Calculates the confidence weights of observations. The formula that determines what percentage of the observation bias is used to correct the predicted state is:

[0084] ; for The covariance matrix of the predicted state at any given time represents the magnitude of the uncertainty in the predicted state; the larger the value, the higher the uncertainty. The observation matrix and the observation model function Correspondingly, the covariance of the state space is mapped to the observation space; The observation noise covariance matrix of satellite positioning data is pre-calibrated based on the performance of the positioning sensor, for example: In real-time dynamic positioning mode: Low noise and high reliability. In single-point positioning: It is noisy and has low trust levels.

[0085] Simply put, The larger the value, the less accurate the prediction. The smaller the value, the more accurate the prediction. The larger the value, the stronger the impact of the observation bias on the state correction.

[0086] The optimal estimate is obtained by updating the state and covariance. The predicted state is then corrected using observation bias. The optimal state estimate at time t is given by the formula:

[0087] ; ; for The estimated optimal state of the system at time t; for The covariance matrix of the optimal state estimate at time step; for The actual measured value of satellite positioning at any given time; The innovation value is the deviation between the observed value and the predicted value, reflecting the error in the predicted state; It is an identity matrix.

[0088] so Eliminated the cumulative drift of the IMU. This updates the uncertainty of the optimal state, which is usually more than... Smaller size, improved precision.

[0089] Satellite positioning data is used to update the position and altitude, eliminating the cumulative drift error of the IMU. If relying solely on the IMU, the position drift could reach tens of meters within an hour. Satellite positioning updates can maintain position and altitude accuracy at the centimeter or meter level over a long period, meeting the needs of precision operations (such as power line inspection and targeted deployment). It also provides an absolute positioning reference, ensuring that the UAV's position estimate is consistent with its actual geographic coordinates, avoiding the problem of accurate relative motion but absolute position deviation.

[0090] Filtering estimation algorithm configuration three: using environmental wind speed data as external disturbance input to compensate for state prediction. This configuration is the core of the filtering algorithm's disturbance resistance. By compensating for the impact of wind disturbances on UAV motion through feedforward, it reduces prediction errors and improves the accuracy of state estimation under complex wind fields.

[0091] Environmental wind speed data comes from wind speed sensors (such as ultrasonic anemometers and pitot tubes), specifically: wind speed vector. In the geodetic coordinate system, reflecting , , Real-time wind speed in the direction of travel. Wind is the main external disturbance to the movement of drones. For example, crosswinds can cause the drone's speed relative to the ground to deviate from the expected trajectory, and gusts can cause attitude fluctuations. If not compensated, these will significantly increase the state bias of the IMU prediction.

[0092] Wind disturbances directly alter the actual motion state of the drone. Therefore, wind speed needs to be introduced as a known external disturbance into the state transition function to correct the prediction model (feedforward compensation), rather than passively correcting it after the deviation occurs. The core formula is:

[0093] ; for The ambient wind speed measurement at that moment; This is the perturbation input matrix, used to map the wind speed vector to the corresponding dimension of the state vector. Wind primarily affects the velocity state, therefore... Only The corresponding row is set to 1, and the other dimensions are set to 0.

[0094] The actual speed of the drone relative to the ground = the speed of the drone itself generated by its own control + the additional speed due to wind disturbance, that is: ; The acceleration measured by the IMU is integrated and needs to be converted to the geodetic coordinate system. The compensated predicted velocity is closer to the actual motion state. Additional speed due to wind disturbance.

[0095] Using environmental wind speed data as an external disturbance input to compensate for state prediction reduces the state prediction error caused by wind disturbance.

[0096] Specifically, the filtering estimation algorithm uses a pre-calibrated process noise covariance matrix and an observation noise covariance matrix, both of which are diagonal matrices determined based on sensor performance.

[0097] By accurately quantifying the uncertainties of the system model and sensor measurements, optimal data fusion weights are provided for the state estimation process. Functionally, the process noise covariance matrix and the observation noise covariance matrix act as tuners for the filter, dynamically balancing the confidence in the predicted values ​​of the system dynamics model with the confidence in the measured values ​​of the sensors. This achieves an optimal trade-off between model error and measurement error, outputting high-precision and high-reliability state estimation results.

[0098] The process noise covariance matrix is ​​usually expressed as Its structure is a diagonal matrix. Among them, each diagonal element For the system state vector The Middle The variance measure of the uncertainty of each state component within a single sampling period due to unmodeled dynamics or external disturbances.

[0099] Its value is directly determined based on the system's dynamic characteristics and the performance of the inertial measurement unit. For example, for the attitude angle obtained from the integral of angular velocity, its process noise variance... Angular random walk coefficients of the gyroscope Relevant, satisfy ,in For sampling time. A larger one The value indicates that the model prediction for this state component is unreliable, and the model will be more inclined to trust the relevant sensor observations when updating.

[0100] The observation noise covariance matrix is ​​usually expressed as Its structure is also a diagonal matrix. Among them, each diagonal element The first in the original measurement data A measure of the measurement noise variance of each observation component.

[0101] The values ​​are strictly calibrated based on the accuracy specifications of each sensor. For example, if the horizontal positioning accuracy of a GPS receiver in real-time dynamic positioning mode is 1.5 cm, then the noise variance of its northward position observation can be set to (0.015)². This value is a quantitative indicator; the smaller the variance, the more reliable the sensor data is considered by the filter. During data fusion, the algorithm will assign a higher weight to the sensor data of that path, making it have a greater impact on the final state estimation result.

[0102] The pre-calibrated diagonal matrix form serves to achieve optimal fusion. By converting sensor performance metrics (such as IMU noise density and GPS positioning accuracy) into precise mathematical parameters, each filter can achieve optimal statistical fusion of multi-source sensor data at the algorithm level. Fixed, pre-calibrated parameters ensure consistent filter behavior across different operating cycles and environments, avoiding convergence problems or instability that may arise from adaptive noise estimation.

[0103] The use of a diagonal matrix in the mathematical form implies that the process noise of each state variable and the observation noise of each sensor are statistically independent. This assumption greatly simplifies the computational complexity of the algorithm and is a reasonable and effective approximation for well-decoupled systems in practical engineering, ensuring the real-time performance of the filter on embedded computing platforms.

[0104] Specifically, a tethered unmanned aerial vehicle (UAV) integrated air-ground intelligent operation system also includes a dynamic resource allocator, configured as follows: Calculate the instantaneous fluctuation range of the power estimate in real time.

[0105] When the fluctuation amplitude exceeds the preset safety threshold, the upper limit of the motor speed command in the fusion control command is adaptively lowered.

[0106] The dynamic resource allocator, as a supervisory adaptive management unit, works by establishing a dynamic constraint relationship between the system's energy supply and the power output of the actuators. By monitoring the power transmission stability of the tethered cable in real time, the dynamic resource allocator proactively limits the maximum power output of the UAV when abnormal fluctuations are detected. Its function is to achieve feedforward safety protection based on energy status, preventing motor overload or flight loss of control under unstable power supply conditions.

[0107] The dynamic resource allocator calculates the instantaneous fluctuation range of the power estimate in real time and receives a sequence of power estimates from the power state observer at a fixed sampling period. The fluctuation amplitude is calculated using the normalized differential method, and the specific formula is as follows:

[0108] ; in, This is the power estimate for the current moment. This is the power estimate from the previous moment. It is a reference power value preset based on the ground power output capacity and cable rated parameters. This formula quantifies the relative rate of change of power per unit time; Instantaneous power fluctuation rate is a dimensionless scalar. It transforms continuous power signals into a quantitative indicator reflecting the magnitude of their changes, providing precise input for subsequent threshold decisions.

[0109] When the fluctuation exceeds a preset safety threshold, the upper limit of the motor speed command output by the airborne controller is adaptively reduced. A safety threshold is preset within the dynamic resource allocator. The system continuously calculates the instantaneous volatility. The instantaneous power fluctuation rate is compared to this threshold. When the instantaneous power fluctuation rate exceeds the safety threshold, adaptive adjustment logic is triggered. Its adjustment mechanism is defined by the following function:

[0110] ; in, This is the default upper limit of the motor speed command. It is the rated speed of the motor. It is a safety factor less than 1. The function ensures that the new upper limit of rotational speed is the smaller of the two values.

[0111] The protection threshold is determined based on the insulation class of the tethered cable, the transient response capability of the power supply, and the dynamic safety margin of the UAV. These are the adjusted constraint parameters sent to the airborne controller. The dynamic resource allocator modifies the control constraints of the model predictive control unit in the airborne controller. To implement this restriction.

[0112] The configuration mechanism of the dynamic resource allocator is essentially a preventative control. By actively reducing the upper limit of motor speed (i.e. thrust), it directly limits the maximum acceleration and power demand that the UAV may reach during periods of power instability, avoiding the risks of abnormal motor operation, controller saturation, or even instability due to insufficient power supply, and ensuring the basic flight safety and controllability of the system under abnormal energy conditions.

[0113] Specifically, the communication link adopts a primary and backup dual-link architecture: The main link is a wireless communication link.

[0114] The backup link is a fiber optic communication link configured using a tethered cable.

[0115] Seamless switching between primary and backup links is achieved through an automatic switching module.

[0116] Furthermore, redundant channels are constructed using heterogeneous physical media, and automatic fault detection and switching mechanisms are employed to ensure continuous transmission of control commands and status data. Functionally, this architecture addresses the risk of system loss of control caused by the interruption of a single communication link due to environmental interference or hardware failure, providing uninterrupted data path assurance for the core control loop.

[0117] The main link is a wireless communication link, which typically uses high-performance wireless modules (such as Wi-Fi 6 modules based on the IEEE 802.11ax standard) and operates in unlicensed industrial, scientific, and medical frequency bands. Its physical layer uses orthogonal frequency division multiplexing (OFDM) technology to transmit data in parallel through multiple orthogonal subcarriers and supports multi-user multiple-input multiple-output (MIMO) technology, enabling high-speed and stable air interface connections with ground stations in complex electromagnetic environments.

[0118] As a regular working channel, it undertakes the majority of the system's data transmission tasks, including uploading UAV status awareness data and downloading mission planning instructions from the ground station. Its high bandwidth characteristics support the transmission of high-throughput information such as environmental awareness point cloud data.

[0119] The backup link is a fiber optic communication link constructed using a tethered cable. This link integrates single-mode or multimode optical fibers within the sheath or reinforcing members of the tethered cable. Optical signals are transmitted within the fiber core based on the principle of total internal reflection, and its physical layer is completely heterogeneous from the main link. The communication terminal uses optical modules to perform electro-optical and photoelectric conversion, modulating and demodulating the electrical signals of the control system with the optical signals in the optical fiber.

[0120] As a highly reliable backup channel, its physical connection characteristics inherently give it strong resistance to electromagnetic interference. When the main link fails due to co-channel interference, physical obstruction, or equipment malfunction, this link can provide a low-latency and stable backup communication path unaffected by the external radio environment, ensuring that the most basic control and status feedback data streams are not interrupted.

[0121] Seamless switching between primary and backup links is achieved through an automatic switching module, which is a daemon process operating at the lower level of the communication protocol stack. It continuously monitors the communication quality indicators of the primary link, primarily including transmission latency, packet loss rate, and link signal strength. This module has preset switching decision thresholds.

[0122] The module periodically sends probe data packets to the ground station and calculates the round-trip time to obtain the latency. At the same time, it counts the number of response timeouts for uplink or downlink data packets to calculate the packet loss rate. When any monitored indicator (e.g., latency consistently higher than 50ms or packet loss rate exceeding 5%) deteriorates to exceed its preset threshold, the switching logic is immediately triggered.

[0123] Once the seamless switching mechanism is triggered, the module does not interrupt the application layer communication session. Instead, it quickly switches the data flow routing path from the wireless network interface to the fiber optic network interface at the network layer or data link layer. This process is achieved through pre-established dual routing tables, virtual IP addresses, or fast port switching technology. For the upper-layer control system, this switching process is transparent and imperceptible, thus ensuring the continuity of the control loop.

[0124] Specifically, a tethered unmanned aerial vehicle (UAV) integrated air-ground intelligent operation system also includes a Lyapunov optimizer, configured to: construct a joint optimization algorithm for system stability constraints and system performance objectives, wherein the stability constraints quantify the cumulative deviation of the system through one or more system state queues.

[0125] The system performance targets are used to characterize the core indicators that the system needs to optimize, including trajectory tracking accuracy, task execution efficiency, and energy consumption economy.

[0126] Based on the power estimate obtained from the power state observer and the real-time length of the system state queue, the weight parameters in the optimization objective function of the model prediction control unit of the airborne controller and the ground controller are dynamically adjusted.

[0127] By solving an optimization problem in each control cycle that aims to minimize the weighted sum of the Lyapunov drift term and the system performance penalty term, the long-term average performance and instantaneous stability of the system are balanced in a dynamic environment.

[0128] A joint optimization algorithm for system stability constraints and performance objectives is constructed. This Lyapunov optimizer first establishes a unified mathematical algorithm that combines the stability requirement of the system not crashing with the performance objective of achieving excellent system performance for holistic consideration. The instability of the system is quantified by creating a virtual queue; for example, a queue representing the backlog of unfinished tasks. If the queue length continues to grow, it means the system is becoming unstable. The core of this algorithm is that it transforms the hard constraint of ensuring stability into a quantifiable and optimization-involved soft objective.

[0129] Based on the power estimate and the system state queue, the weight parameters in the optimization objective function of the airborne controller and the ground controller are dynamically adjusted. The optimizer monitors the power estimate from the tethered cable and the system state queue, which reflects system stability, in real time. Based on this real-time data, it dynamically issues parameter tuning instructions to the underlying airborne and ground controllers.

[0130] For example, when it detects drastic fluctuations in power supply, it instructs the controller to penalize larger control actions in the calculation, that is, to increase the weighting parameter of the control quantity, so that both the first and second control commands are more inclined to save energy and ensure stability. Conversely, when the power is sufficient and the system queue is stable, it instructs the controller to penalize tracking errors more, that is, to increase the weighting parameter of the state error, so that the UAV can track the target trajectory more accurately.

[0131] By minimizing the weighted sum of Lyapunov drift and system performance penalty terms, the average performance and dynamic stability of the system are balanced. In each control cycle, the Lyapunov optimizer operates to find an optimal decision that minimizes a specific cost function.

[0132] This cost function consists of two parts: the first is the Lyapunov drift, which represents the expected change in the system's instability; minimizing it means striving to bring the system towards a stable state. The second part is the system performance penalty, which represents the degree of poor performance the system would have under the current decision. A tradeoff factor adjusts the weights of these two parts. By continuously minimizing this weighted sum, the optimizer ensures that the system does not sacrifice long-term stability (such as running out of power) in pursuit of short-term performance optimization (such as flying fast), nor is it overly conservative in pursuit of absolute stability, thus achieving dynamic equilibrium and global optimum in the long run.

[0133] Based on the above-mentioned tethered UAV integrated air-to-ground intelligent operation system, this technical solution also provides a tethered UAV integrated air-to-ground intelligent operation method, including the following steps: The voltage and current signals of the moored cable are collected in real time by a power state observer, and the power estimate is calculated.

[0134] The power state observer continuously monitors the electrical parameters of the moored cable using a high-precision metering chip, based on a physical model of cable power transmission. The observer multiplies the acquired instantaneous voltage and current values ​​to obtain the apparent power, then subtracts the heat loss power calculated based on the cable's inherent resistance and the square of the current, ultimately outputting an accurate estimate of the net transmitted power. This process transforms the raw voltage and current signals into key indicators reflecting the real-time energy supply status, providing an accurate energy budget basis for subsequent intelligent decision-making.

[0135] The drone's pose information and environmental disturbance data are acquired in real time through a multi-sensor fusion positioning module.

[0136] The multi-sensor fusion positioning module, acting as the system's perception center, receives raw data from heterogeneous sensors such as the inertial measurement unit, satellite positioning receiver, and ultrasonic anemometer. This module employs an advanced filtering estimation algorithm, which overcomes the limitations of a single sensor through data complementarity: it uses the high-frequency characteristics of inertial data to predict dynamics, leverages the absolute accuracy of satellite data to correct accumulated errors, and uses wind speed measurement as a known disturbance input. Through this fusion processing, the module ultimately outputs UAV pose information and environmental disturbance data that combine high-frequency update rate, high accuracy, and strong anti-interference capabilities.

[0137] Based on the power estimate, pose information, and environmental disturbance data, the airborne controller generates a first control command in a first control cycle.

[0138] The airborne controller, acting as the local decision-making core, operates continuously in short cycles down to the millisecond level. Its working principle involves inputting received power estimates, real-time pose information, and environmental disturbance data into a Model Predictive Control (MPC) algorithm. Within each control cycle, this algorithm comprehensively considers the current state, future dynamics, operational constraints, and available power budget to quickly solve for the optimal control sequence within a finite time domain. Ultimately, it outputs the first control command, designed for immediate execution, ensuring the UAV makes agile and precise instantaneous adjustments to its attitude and trajectory.

[0139] Based on the power estimate, pose information, and environmental disturbance data, a second control command is generated by the ground controller in a second control cycle, wherein the first control cycle is shorter than the second control cycle.

[0140] As a higher-level decision-making unit, the ground controller has a significantly longer operating cycle than the airborne controller. Its working principle involves utilizing more abundant computing resources to process the same state information on a slower timescale. The optimization algorithms executed by the ground controller focus on long-term performance. Its task is to generate strategically significant second control commands based on a comprehensive consideration of overall mission progress, energy consumption strategies, and global path planning. These commands do not directly drive the motors but rather provide high-level target setting and behavioral guidance for the airborne, fast controller.

[0141] The first control command and the second control command are fused together to generate the final control command that drives the UAV to fly.

[0142] This technical solution employs a dynamic weighted fusion strategy to process control commands from different levels. The strategy works by adaptively adjusting the contribution weight of the two levels of commands in the final output based on real-time system status, particularly the stability of power supply. When the system is running smoothly and energy is sufficient, the fusion strategy prioritizes the second control command, which reflects global optimization. When drastic power fluctuations are detected or rapid attitude stabilization is required, the weight of the first control command, which reflects local reaction speed, is significantly increased. Through this intelligent fusion, the final generated fused control command can both fulfill long-term mission objectives and ensure instantaneous flight safety and stability.

[0143] Specifically, the fault-tolerant control steps include: real-time monitoring of the transmission delay of the communication link used for timestamp synchronization.

[0144] When the transmission delay exceeds the preset delay threshold, the drone terminal switches to local capacity control mode and triggers an alarm signal.

[0145] The fault-tolerant control process involves real-time monitoring of the communication link's transmission delay, establishing a continuous perception mechanism for the link's performance. This works by embedding a precise transmission timestamp into each transmitted data packet. The receiving end calculates the round-trip time of the signal in the transmission medium by comparing the reception time with the timestamp within the packet. The system uses a high-precision clock source to provide a unified timestamp for all data packets, ensuring the accuracy of delay calculations. This mechanism continuously tracks changes in the communication link's quality, providing real-time assessments of the system's communication status and serving as a fundamental condition for triggering subsequent fault-tolerant operations.

[0146] The fault-tolerant control mechanism that determines when the transmission delay exceeds a preset delay threshold forms the core of the system's fault identification and decision-making. The system compares the real-time calculated transmission delay value with a pre-set safety threshold, which is determined based on the stability requirements of the control system, algorithm convergence time, and task criticality. When the system detects that the delay continuously exceeds this threshold and reaches a preset confirmation period, it determines that the current communication link can no longer meet the basic requirements for reliable control. This judgment logic employs necessary fault-tolerant design to avoid false triggers caused by instantaneous fluctuations, while ensuring a timely response when actual communication quality deterioration occurs.

[0147] The fault-tolerant control process involves switching the UAV to local fault-tolerant mode. This step is the response action of fault-tolerant control. Upon confirming a communication failure, the system immediately switches control from a global control mode relying on ground-to-ground collaboration to a fully autonomous local fault-tolerant mode. During the switch, the system uses the last received valid fused control command as a reference, combined with real-time status information provided by raw data from heterogeneous sensors such as the inertial measurement unit, satellite positioning receiver, and ultrasonic anemometer, and employs a specially designed local control algorithm to maintain the UAV's basic flight functions. The local fault-tolerant mode prioritizes flight safety, maintaining stable system operation even without ground support by simplifying control logic and limiting maneuverability, thus creating the necessary conditions for subsequent recovery or rescue operations.

[0148] The fault-tolerant control steps activate an alarm signal warning mechanism, establishing a comprehensive human-machine collaborative warning system. Simultaneously with the trigger mode switch, the system sends a clear fault signal to ground operators via a multi-modal alarm channel. Visual alarm devices employ a specific frequency flashing pattern, while audible alarm devices emit regular alert sounds. Simultaneously, the ground monitoring interface displays detailed fault types and handling suggestions. This multi-layered alarm mechanism ensures effective attention from operators under various environmental conditions, providing ample information support for timely understanding of system status and intervention measures, thus forming a complete safety assurance loop.

[0149] Specifically, the step of fusing the first control command and the second control command includes: The joint optimization algorithm constructed using the Lyapunov optimizer dynamically adjusts the fusion weights of the first and second control commands based on the real-time fluctuations of the power estimate. When the fluctuation range of the power estimate is within the first preset range, the weight ratio of the second control command is increased; When the fluctuation range of the power estimate is within the second preset range, the weight ratio of the first control command is increased.

[0150] Furthermore, the instruction fusion strategy establishes a fusion mechanism based on an optimization algorithm. A dynamic weight allocation algorithm is built based on Lyapunov optimization theory, using the real-time fluctuations of the power estimate as the core decision variable. This algorithm quantifies the deviation between the system state and the ideal target by constructing a virtual queue, and aims to minimize the weighted sum of Lyapunov drift and performance penalty terms. This mechanism transforms abstract system stability and performance requirements into concrete mathematical optimization problems, providing a theoretical basis and execution algorithm for intelligent weight allocation, ensuring that the fusion process satisfies both mathematical optimality and practical engineering needs.

[0151] The implementation process of dynamically adjusting fusion weights based on power fluctuations in the command fusion strategy involves the system continuously monitoring the instantaneous fluctuation amplitude of the power estimate and normalizing it into a standardized volatility index. This volatility is obtained by calculating the percentage of power change relative to rated power per unit time, providing precise quantitative input for weight adjustment. Based on real-time volatility data, the system dynamically calculates the weight coefficients of the first and second control commands in the final output using a preset mapping function. This design enables the system to automatically adjust its dependence on rapid response and global optimization capabilities according to the stability of energy supply, achieving online adaptive control strategy.

[0152] The command fusion strategy includes a decision logic that increases the weight of the second control command when fluctuations are within a first preset range. This conditional branch corresponds to the system's optimization strategy under stable energy conditions. When the fluctuation range of the power estimate is detected to be within the first preset range, it indicates that the power transmission of the tethered cable is stable and the energy supply is sufficient and reliable. At this time, the system increases the weight of the second control command, making the globally optimized command generated by the ground controller dominate the final control output. This decision enables the UAV to better execute the mission commands calculated through global optimization, fully utilize stable energy conditions to achieve optimal trajectory tracking, energy efficiency management, and mission scheduling, fully leverage the powerful computing capabilities of the ground station, and improve the overall operational efficiency of the system.

[0153] The command fusion strategy includes a decision logic that increases the weight of the first control command when fluctuations fall within a second preset range. This condition corresponds to the system's safety assurance strategy under unstable energy conditions. When the fluctuation range of the power estimate enters the second preset range, it indicates abnormal fluctuations in cable power transmission and uncertainty in energy supply. At this time, the system increases the weight of the first control command, allowing the fast-response commands generated by the onboard controller to play a greater role in the final control output. This decision ensures that the UAV prioritizes maintaining basic flight stability under unstable energy conditions. By strengthening the role of the local fast control loop, it promptly suppresses attitude instability and trajectory deviations that may be caused by power fluctuations, providing reliable assurance for the safe operation of the system under abnormal conditions.

[0154] In one specific embodiment, at 8:00 AM, inspector Xiao Wang starts the system. The drone slowly rises from the ground station, and the tethered cable begins supplying power and transmitting data. Today's task is to inspect a 20-kilometer-long high-voltage line.

[0155] As soon as the drone took off, all parts of the system began to work together: Power monitoring: The onboard power status observer constantly calculates the actual power transmitted by the cable to ensure a stable energy supply.

[0156] Precise positioning: The multi-sensor fusion positioning module combines the absolute accuracy of GPS with the sensitive response of IMU, allowing the drone to always know its precise location with an error of only a few centimeters.

[0157] Intelligent decision-making: The airborne controller generates the first control command every 10 milliseconds to adjust the flight attitude; the ground controller generates the second control command every 100 milliseconds to plan the optimal inspection route. The system merges the two into a fused control command and then executes it.

[0158] When the drone reached the 5-kilometer mark during its inspection, it suddenly encountered a crosswind: the anemometer immediately detected the change in wind force, the filtering estimation algorithm quickly compensated for the wind's influence, and the onboard controller immediately adjusted the motor speed to stabilize the drone.

[0159] Around midday, the strong sunlight caused interference with the wireless signal. The communication link automatically detected the increased latency and seamlessly switched to the fiber optic backup link within 50 milliseconds, ensuring the inspection work was not affected in any way.

[0160] At 2 PM, the startup of large equipment at a nearby factory caused power fluctuations. The power status monitor immediately detected that the power fluctuation exceeded 15%, and the dynamic resource allocator intervened immediately, appropriately reducing the upper limit of motor power. The drone automatically reduced its maneuverability to prioritize stable flight. Once the power supply stabilized, the system automatically restored normal power.

[0161] During 8 hours of continuous operation, the Lyapunov optimizer dynamically adjusted the control strategy based on real-time data, ensuring both flight safety and improved inspection efficiency.

[0162] Ultimately, the drone successfully completed its inspection mission.

[0163] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A tethered unmanned aerial vehicle air-ground integrated intelligent operation system, characterized in that, Includes the drone terminal, the ground station, and the tethered cable connecting the drone terminal and the ground station; The UAV terminal is equipped with an onboard controller, and the ground station is equipped with a ground controller; The multi-sensor fusion positioning module, installed on the drone, is used to provide the status information required for the drone's flight. Based on the fact that the control cycle of the airborne controller is shorter than that of the ground controller, a distributed model prediction control unit is constructed. The UAV terminal is also equipped with a power status observer, which is used to calculate the power estimate based on the collected voltage and current signals of the mooring cable and the resistance parameters of the mooring cable. The airborne controller generates a first control command by performing rapid response calculations based on the status information in a first control cycle, and the ground controller generates a second control command by performing global optimization calculations based on the received power estimate and status information in a second control cycle. The system also includes a Lyapunov optimizer, which dynamically adjusts the weight parameters in the objective function of the model prediction control unit of the airborne controller and the ground controller based on the power estimate obtained from the power state observer and the real-time length of the system state queue characterizing the cumulative deviation of the system. When fusing the first control command and the second control command, the fusion weight of the first control command and the second control command is dynamically adjusted according to the real-time fluctuation of the power estimate. Specifically, when the fluctuation range of the power estimate is within a first preset range, the weight ratio of the second control command is increased, and when the fluctuation range of the power estimate is within a second preset range, the weight ratio of the first control command is increased. A fused control command to drive the UAV to fly is generated and issued.

2. The tethered unmanned aerial vehicle air-ground integrated intelligent operation system according to claim 1, characterized in that, The model prediction control unit is configured as follows: Compressed prediction and control time domains are employed; A differentiated weight matrix is ​​used to perform weighted optimization of the system state vector error and control quantity; The weight assigned to position state error is greater than the weight assigned to attitude state error.

3. The tethered unmanned aerial vehicle (UAV) integrated air-ground intelligent operation system according to claim 1, characterized in that, It also includes a multi-sensor fusion positioning module, which uses a filtering estimation algorithm to fuse inertial measurement data, satellite positioning data and environmental wind speed data to obtain the state information containing pose information and environmental disturbance data; The filtering estimation algorithm is configured as follows: The inertial measurement data is used as the basis for system state prediction; The position and altitude of the UAV are observed and updated using the satellite positioning data. The environmental wind speed data is used as an external disturbance input to compensate for the system state prediction.

4. The tethered unmanned aerial vehicle air-ground integrated intelligent operation system according to claim 3, characterized in that, The filtering estimation algorithm uses a pre-calibrated process noise covariance matrix and observation noise covariance matrix, both of which are diagonal matrices determined based on sensor performance.

5. The tethered unmanned aerial vehicle air-ground integrated intelligent operation system according to claim 1, characterized in that, It also includes a dynamic resource allocator, configured as follows: Real-time calculation of the instantaneous fluctuation range of power estimates; When the fluctuation amplitude exceeds the preset safety threshold, the upper limit of the motor speed command in the fusion control command is adaptively lowered.

6. The tethered unmanned aerial vehicle air-ground integrated intelligent operation system according to claim 1, characterized in that, The communication link adopts a primary and backup dual-link architecture: The main link is a wireless communication link; The backup link is a fiber optic communication link configured using a tethered cable. Seamless switching between primary and backup links is achieved through an automatic switching module.

7. A tethered unmanned aerial vehicle air-ground integrated intelligent operation method, characterized in that, Includes the following steps: The voltage and current signals of the moored cable are collected in real time by a power state observer, and the power estimate is calculated by combining the resistance parameters of the moored cable. The drone's pose information and environmental disturbance data are acquired in real time through a multi-sensor fusion positioning module. Based on the power estimate, pose information and environmental disturbance data, the airborne controller generates a first control command in a first control cycle. Based on the power estimate, pose information and environmental disturbance data, a second control command is generated by the ground controller in a second control cycle, wherein the first control cycle is shorter than the second control cycle. A joint optimization algorithm constructed using the Lyapunov optimizer dynamically adjusts the weight parameters in the objective function of the airborne controller and the ground controller based on the power estimate and the real-time length of the system state queue characterizing the cumulative deviation of the system. When fusing the first control command and the second control command, the fusion weight of the first control command and the second control command is dynamically adjusted according to the real-time fluctuation of the power estimate. Specifically, when the fluctuation range of the power estimate is within a first preset range, the weight ratio of the second control command is increased, and when the fluctuation range of the power estimate is within a second preset range, the weight ratio of the first control command is increased, thereby generating a fused control command to drive the UAV to fly.

8. The tethered unmanned aerial vehicle air-ground integrated intelligent operation method according to claim 7, characterized in that, It also includes fault-tolerant control steps: Real-time monitoring of the transmission latency of the communication link for timestamp synchronization; When the transmission delay exceeds the preset delay threshold, the drone terminal switches to local capacity control mode and triggers an alarm signal.

Images